An Adaptive Method for Dynamic Audience Size Estimation in Multicast

摘要

We develop an end-to-end protocol for real-time estimation of the size of dynamic multicast groups. Unlike previously proposed methods our approach alleviates feedback implosion in a dynamic setting, and is scalable to large groups. The protocol is based on probabilistic polling combined with adaptive feedback control, and the use of a time-dependent Wiener filter to enhance estimation accuracy. We examine the performance of our protocol through simulations for multicast groups with up to 10,000 members, and different scenarios of group membership dynamics. Our simulation studies show that the method is capable of tracking, in a scalable manner, the size of dynamic multicast groups with high accuracy in the face of large dynamic variations.